White Flight: The Effect of Minority Presence on Post World War II Suburbanization
Eric Bickford, University of California-Berkeley
Immediately following World War II, the United States underwent an urban transformation. Kenneth T. Jackson, citing a Bureau of Labor Statistics survey for home building in the six largest metropolitan areas for 1946-47, notes that over 62% of all home c onstruction occurred in suburban areas. US Census data show that for the period 1950-1960, central city populations in the largest 25 SMSAs increased by just over 3%, while total suburban populations increased by well over 60%.1 While the total populatio ns of the nation's largest central cities stagnated, the number of blacks in the central cities increased substantially, implying an urban depopulation by whites who migrated to the suburban fringes. The numbers for all the nation's metropolitan areas ar e somewhat less striking. However, even in this broadly based sample, the distribution of the population shifted: the suburbs remained largely white despite tremendous growth and the cities became increasingly nonwhite.
However, from this evidence alon e one cannot necessarily conclude that the black/white - city/suburb dichotomy arose purely on racial grounds. Cities in the United States tend to follow a pattern of neighborhood succession in which wealthy residents abandon smaller, older units of hous ing in favor of larger, newly constructed units. As the existing housing stock in older neighborhoods depreciates with age, it becomes occupied by lower-income households; a neighborhood transformation takes place in which older, central neighb orhoods decline while new affluent communities form on the urban fringes. Since blacks historically have had lower incomes than whites, as shown in Figure 1, the supposed post-war "white flight" could be attributed to simple neighborhood succession. Whi te households, with their higher incomes, were able to locate in the new, more expensive housing in these new neighborhoods, while nonwhites, with lower incomes, increasingly occupied the older, less desirable areas near urban centers. These new neighbor hoods simply happened to lie outside the political boundaries of the large central city, and the normal pattern of neighborhood succession was given a new name - suburbanization.
This reasoning is somewhat tempting. After World War II, housing became affordable to more families because of new mass production techniques, government financing, relatively low interest rates,2 and high real wages. A family with the median white post war annual income of $3000 could handily have afforded the no-down-payment mortgage for a $7,990 Cape Cod production house in William Levitt's famous Levittown subdivision.3 The median nonwhite income, just over $1500, would be insufficient. Upon caref ul examination of the evidence, this reasoning fails. First, at least a small fraction of the nonwhite population would presumably have earned sufficient income to locate in the new suburbs. If the racial distribution of the population were due simply t o differences in income, then higher income nonwhites, able to afford the newly produced housing, would have located in the new developments alongside whites. In the case of Levittown, NY, however, not a single one of the original 82,000 residents was bl ack.4
Reynolds Farley finds a tendency toward racial segregation after adjusting for income and educational differences between races. Regardless of socioeconomic status, whites tend to live in the suburban ring, while blacks disproportionately inhabit the cit ies. Finding proportionally fewer college-educated blacks than grammar-school-educated whites in the suburbs, Farley concludes that the absence of blacks from the suburbs cannot be explained entirely by socioeconomic status.5
One can explain the apparent racial segregation in one of two basic ways: the "pull" hypothesis and the "push" hypothesis. The "pull" hypothesis asserts that, following World War II, a combination of federal housing programs, highway expenditur es, and mass production techniques combined to make suburbanization possible. It assumes that there is an intrinsic lure to suburban residence and that when the above events occurred in postwar America, suburbanization resulted as thousands of househ olds were "pulled" out from the city centers by the amenities of the suburbs. This hypothesis ex-plains racial segregation as a function of the extent to which nonwhites were denied access to the means of suburban home purchase, through discrimination by government programs and "redlining" by lending in-stitutions and realtors.
The opposite theory explaining the nature of postwar suburbanization, the "push" hypothesis, implies that the movement to the suburbs was substantially affected by changing social conditions in the central cities. Proponents of t his hypothesis charge that white urbanites, in response to an increasing population of members of undesirable minority groups, chose to migrate outward to achieve a more homogeneous lifestyle in the suburbs. They argue that the postwar trend toward subur ban development is substantially a result of the large urban influx of ethnic minorities, largely Southern blacks, who migrated to the cities seeking employment in wartime factories.
In this paper, I attempt to examine the "push" hypothesis of suburbanization. Specifically, to what extent did the growing nonwhite population of the central cities motivate middle-income whites to establish residence in the homogenous suburban ring?
To answer this question, I first examine the rhetorical arguments which underlie the reasoning of the "push" hypothesis. Upon establishing it as a plausible explanation, I analyze population data for the largest metropolitan areas in the United States to confirm or deny a correlation between the "push" factors and the deconcentration of metropolitan population.
Arguments in Favor
There are three plausible "push" explanations for this suburban flight: the distribution of key government services, primarily public education, the role of status in neighborhood formation, and the nature of political behavior of various socioeconomic cl asses. First is the notion of shared institutions and governmental services, especially publicly funded schools. The Supreme Court's Brown v. Board of Education decision in 1954 desegregated school districts nationwide. Proponents of the public-school explanation argue that following the Brown decision, schoolchildren who once attended classes only with the homogenous population of their urban neighborhood enclaves were suddenly placed in classrooms alongside children from less desirable urban areas. By locating outside the jurisdiction of the central city and its corresponding school district, white families were able to circumvent the Brown decision by placing their children in districts with a majority white population. Using the separate politica l institutions of the newly forming suburban communities, whites were able to escape the rising tide of integration and establish their own schools, libraries, police, and parks.
There is evidence from the real estate industry to substantiate this claim. Anthony Downs, a senior real estate consultant testifying before the Senate Select Committee on Equal Educational Opportunity, argued that "for most middle-class Americans with s chool-age children, choosing a place to live is greatly influenced by the nature of the schools serving the various neighborhoods available to them" and that "middle-class Americans . . . typically judge the quality of any school mainly in the kind of fam ilies whose children are predominant in its classrooms."6 By this reasoning, white families discriminate against nonwhite schools not necessarily out of blatant prejudice against nonwhites, but because they perceive schools enrolling unusually large numb ers of nonwhites to be suspect or inferior and wish to place their children in more ethnically homogenous schools which they perceive to be superior. This phenomenon can be observed in the case of Park Forest, Illinois, which, prior to 1970, was a largely white, middle-class community south of Chicago. When a black housing development was built on a tract of land outside the boundaries of Park Forest and children from the development began attending local schools, whi te residents began selling their homes, fearing a drop in land values. As land values dropped, more blacks began to occupy the once-white community and a majority-white area was transformed in the span of a few years.7
The second explanation, that of status, depends upon the importance of the perception of status to the middle-class population. This reasoning assumes that the paranoid, status-conscious middle-class resident selects a neighborhood based on the socia l status it conveys to the rest of society. While the individual may not necessarily be a racist, he acknowledges the correlation between low status and membership in an ethnic minority group that is determined by the rest of the racist socie ty. He chooses not to locate in a neighborhood of nonwhites because to do so would equate his own status with that of the existing low-status residents. This is ex-plicitly expressed by postwar author Gerald Abrams who writes, "It is no longer the type of house but the type of neighborhood which reflects social standing . . . Fine looking homes in Chicago may be still as fine looking but are considered blighted when Negroes or other minorities live in the nei ghborhood."8
The third possibility is that of political behavior. This explanation divides individual assumptions about political institutions into two camps: the public-regarding and private-regarding. This theory, advanced by Banfield and Wilson in City Politics, suggests that the older WASP members of American society are descended from a Colonial tradition of a classless society and enlightened self-government by and for the people. These public-regarding individuals favor nonpartisan local government in which concerned citizens act altruistically for the public good. Conversely, "private minded" blacks and the newer immigrants from southern Europe have a fundamental distrust of the political process and view politics as a brokerage of private interests by var ious competing groups. According to Banfield and Wilson, private-regarding citizens identify strongly with members of their own group and see the governmental process as a means for advancing their own interests, while public-regarding citizens identify with the community at large and see government as a means of achieving good for all.9 The push hypothesis asserts that public-regarding white inhabitants of the central cities, dismayed over the machine politics and contentiousness of special interests w hich proliferated as a result of the expanding private-regarding (nonwhite) population, migrated to the suburbs to re-establish "good government". Prewitt and Eulau, examining the voting behavior of 80 San Francisco Bay area cities, find that low-status communities tend to have high voter turnouts and high rates of eviction of public officials, while high-status communities tend to have relatively low voter turnouts and high rates of incumbency.10 This seems to confirm the suspicion of government by low -status, private-regarding individuals, and the relative comfort with government by high-status, public-regarding individuals.
Thomas Guterbock, in "The Push Hypothesis", develops a measure of urban deconcentration in which he assumes that the geographic distribution of urban population is exponential: population density is greatest at the center and declines exponentially in all directions from the center out to the edge of the metropolitan area. He represents the density at a certain point in the city as Dx= ae-bx, where x represents the distance from a city's center and a and b are constants. For a group of 19 S tandard Metropolitan Statistical Areas (SMSAs), he calculates the parameters a and b based on available data on population size and land area. From these parameters a and b, he is able to estimate a density gradient for each SMSA, arrives at levels of concentration for the years 1950, 1960, and 1970 and computes a rate of relative deconcentration for each SMSA.11
Guterbock's method suffers from two basic flaws. First, it is applicable only to monocentric, circle or pie-shaped SMSAs. This eliminates areas like San Francisco, which have complex geographical limitations that do not conform to his simple geometric d efinition. Second, it cannot account for areas with more than one central city, like New York-Newark, Los Angeles-Long Beach, or Dallas-Fort Worth. Third, it assumes that the push hypothesis refers primarily to density and distance, rather than politica l boundary. If one accepts that the decision to migrate to the suburbs is motivated by the promise of independent government, then political location, as well as geographic location, is also significant.
As a measure of urban concentration that incorporates both population density and "political density", I used the measure (CITY/SUB)2, where CITY equals the central city population in a given year, and SUB equals the suburban fringe population in the same year, for a given SMSA. Although actual population densities (i.e., population per square mile) vary significantly across SMSAs - e. g., the city of New York is physically much denser than the city of Los Angeles - this ratio can still provide relative comparisons for political density for a given SMSA. The higher the density value, the greater the preference a given metropolitan population has for its central city, and presumably, the lower the central city's corresponding disutility. This term is sq uared to attempt to incorporate the concept of physical density. Taking Guterbock's assumption that density increases exponentially as one moves toward the center of an urban area and that the central city or cities generally lie at the center of a metropolitan area with the suburbs forming an outer ring, the ratio (CITY/SUB)2 will also crudely approximate an area's population density gradient. As the (CITY/SUB) ratio declines, the population becomes more geographically and politically fr agmented.
To study the effects of the push factors on urban concentration I calculated concentration ratios for the 25 largest SMSAs in 1960, excluding any SMSAs, such as Paterson, NJ, which are classified as part of a consolidated area within another, larger SMSA. Four of the SMSAs, Dallas, Houston, Atlanta, and Miami, were excluded because their central cities annexed heavily over the period, resulting in a high level of central city concentration which may more strongly represent local governmental fiat rather than individual residential choice.
As explanatory variables, I used the following. ĂNW(40-50), the absolute change in the number of nonwhite city residents for the years 1940-50, measures the push effect of the increasing numbers of nonwhites in the central city. NW50 represents the perc entage of the city population that is composed of nonwhites in 1950. LOGMSA50, the natural logarithm of the total SMSA population for 1950, represents the effects of relative metropolitan size on concentration. INC50 and INC60 are the real median family incomes, deflated by the CPI for each SMSA as given by the Bureau of Labor Statistics for 1950 and 1960.
For Equation (1), shown in Table 1, I regress CON60 on the explanatory variables NW50, ĂNW(40-50), LOGMSA50, LOGINC50, and LOGINC60.
Only the variables NW50, ĂNW(40-50), and LOGMSA50 are statistically significant. As predicted by the push hypothesis, there is a negative correlation between both the relative size of the nonwhite central city population (NW50) and the level of concentra tion (CON60). More surprisingly, the correlation with the absolute increase in nonwhite population [ĂNW(40-50)] is positive. This implies the possibility of two simultaneous effects: while a high percentage of central-city nonwhites might correspond with a white exodus to the suburbs, a large sustained influx of nonwhites to the central cities helps to maintain relative density by populating the central cities. Put another way, a city in 1950 with a relatively high percentage but relatively low in-migration of nonwhites tends to have a lower concentration score, while a city with a low percentage but a large influx of nonwhites tends to have a higher concentration score.
To test this hypothesis, one might then calculate a measure of white concentration, similar to the one used above for the total population, and regress it on the above explanatory variables. This would likely produce a negative coefficients for the ĂNW(4 0-50) variables, but only as a matter of mathematical identity. Since the suburban population remained 95% white over the period, a correlation between nonwhite in-migration and white deconcentration would represent the inverse relation between white and nonwhite population which exists, by definition, within the central city.
Modifying the equation to eliminate the statistically insignificant income terms, LOGINC50 and LOGINC60, does not produce substantial changes, as seen in Equation (2). Figure 2, the residual plot of Equation (1), shows the predictive value of this model. The relative importance of each of the terms is shown in Table 2, the plot of the standardized regression coefficients. Ultimately, NW50 accounts for less than 18% of the predicted variation within the model and less than 11% of the total variation in CO N60, the (CITY/SUB)2 population ratio for 1960.
The model tends to support the assertions of the push hypothesis. While the effect of nonwhite population is relatively minor, it is nonetheless present. The reasoning behind the push hypothesis is compelling, but it clearly does not fully explain postw ar migratory patterns, nor does it address the assertions made by the pull hypothesis. Indeed, it is entirely possible that nonwhite concentration is simply a result of large-scale white suburbanization, not the cause. Even if we interpret the model as causal, the effects of the push factors are relatively small.
Additionally, the model is incapable of controlling completely for metropolitan growth. Because of the definition of SMSA population (City + Suburb = SMSA), it is impossible to include the 1960 SMSA population as an explanatory variable, thereby controll ing for urban growth. It may be that many areas deconcentrated simply because the central cities failed to annex while large numbers of people flowed into the large SMSAs.
These flaws aside, there is merit to the push hypothesis at some level. While the effect may or may not be huge, it likely did exist. In the face of increasing fragmentation between neighboring municipalities, it provides a framework for understanding t he current conflict between cities and suburbs as a function of the underlying reasons for which they exist in the first place. By accepting the idea that elements of racism have played a part in community formation, we understand that we must first addr ess those lingering elements. Only then can communities trust each other sufficiently to cooperate and tackle the more mundane regional planning problems of transportation, air quality, taxation, and public services.
Jackson, Kenneth T., The Crabgrass Frontier, New York: Oxford Univ. Press, 1985, p. 238.
2Ibid., p. 241.
3Ibid., p. 236.
4Ibid., p. 241.
5Farley, Reynolds, "Components of Suburban Population Growth", in The Changing Face of the Suburbs, ed. by Schwartz, Barry, Chicago: Univ. of Chicago Press, 1976, p. 19.
6Downs, Anthony, "Suburban Housing: A Program for Expanded Opportunities", Real Estate Review, Spring 1971, Vol. 1, No. 1, Boston: Warren, Gorham, & Lamont, Inc., p. 4.
7Berry, Brian J. L, Goodwin, Carole A., Lake, Robert W., and Smith, Katherine B., "Attitudes toward Integration: the Role of Status in Community Response to Racial Change", in The Changing Face of the Suburbs, ed. by Schwartz, Barry, Chicago: Univ. of Chi cago Press, 1976, p. 240.
8Abrams, Gerald, Forbidden Neighbors, New York: Harper & Bros., 1955, p. 139.
9Banfield, Edward, and Wilson, James Q., City Politics, Cambridge: Harvard Univ. Press, 1963, pp. 40-41.
10Prewitt, Kenneth and Eulau, Heinz, "Social Bias in Leadership Selection, Political Recruitment, and Electoral Context", Journal of Politics, Vol 33, p. 306.
11Guterbock, Thomas M., "The Push Hypothesis: Minority Presence, Crime, and Urban Deconcentration", in The Changing Face of the Suburbs, ed. by Schwartz, Barry, Chicago: Univ. of Chicago Press, 1976, pp. 140-146.
12The total population and concentration scores were calculated from metropolitan population data in U. S. Bureau of the Census 1960, Census of Population, Vol I, Table 33. The nonwhite city population figures for 1940 and 1950 are located in U. S. Burea u of the Census, 1940, Vol II, part 1, Table 86, and 1950, Vol II, Part 1, table 50, respectively. The income data for 1950 and 1960 are taken from U. S. Bureau of the Census, Census of Population, Vol I, part 1, table 148 and are deflated by the CPI's f or all prices given for the respective cities in U. S. Bureau of the Census, Statistical Abstract of the United States, Washington, 1974, No. 667.